Results 1 to 2 of 2

Thread: Showing existence of an operator

  1. #1
    Senior Member Dinkydoe's Avatar
    Dec 2009

    Showing existence of an operator

    I've been given the following question:

    Let $\displaystyle (\mathcal{H},(,))$ be a real Hilbert space with an orthonormal basis $\displaystyle \left\{e_n\right\}_{n=1}^{\infty} $. Let $\displaystyle F:[0,1]\to \mathcal{H}$ be continuous. Show that there exists a positive self-adjoint operator $\displaystyle T\in B(\mathcal{H})$ such that $\displaystyle (Tx,y) = \int_0^1(F(t),x)(F(t),y)dt$ for all $\displaystyle x,y\in \mathcal{H}$.

    Show moreover that $\displaystyle T$ is compact.

    (you may use that $\displaystyle \int_0^1\sum_{n=N+1}^{\infty}|(F(t),e_n)|^2dt\to 0$ as $\displaystyle N\to\infty$)

    I don't quite understand what must be shown here exactly. Must I find such a operator T explicitly and then show it has the desired property. If so, how? I don't quite see where the hint is useful, I guess it's meant for showing compactness of T.

    Any hint or push in the right direction is very much appreciated.
    Follow Math Help Forum on Facebook and Google+

  2. #2
    MHF Contributor
    Opalg's Avatar
    Aug 2007
    Leeds, UK
    [This is not a solution, or even a hint, but it might give you a push in the right direction.]

    To get an idea of what is going on here, start by looking at the very simplest situation, when the function F is constant, say $\displaystyle F(t) = z\in\mathcal{H}$. Then $\displaystyle (Tx,y) = (z,x)(z,y)$, so $\displaystyle Tx = (z,x)z$. This is a rank-one operator (you may perhaps have seen the notation $\displaystyle T = z\otimes z$ for such an operator).

    Next, take a slightly more complicated situation, where $\displaystyle F(t) = z_j$ for $\displaystyle (j-1)/n\leqslant z_j<j/n\;(1\leqslant j\leqslant n)$. (Admittedly, that function F is not continuous, but the idea is to use it as a step function giving a Riemann approximation to a continuous function.) Then $\displaystyle (Tx,y) = \frac1n\sum_{j=1}^n(z_j,x)(z_j,y)$. So $\displaystyle Tx = \frac1n\sum_{j=1}^n(z_j,x)z_j$, or $\displaystyle T= \frac1n\sum_{j=1}^nz_j\otimes z_j$. This is again a finite-rank operator.

    Now use the Riemann approximation idea. Given a continuous function $\displaystyle F:[0,1]\to\mathcal{H}$, let $\displaystyle z_j = F(j/n)\;(1\leqslant j\leqslant n)$, and try to show that as $\displaystyle n\to\infty$ the operators $\displaystyle \frac1n\sum_{j=1}^nz_j\otimes z_j$ defined as in the previous paragraph converge in norm to T. Then T will be a norm limit of finite-rank operators, hence compact.
    Follow Math Help Forum on Facebook and Google+

Similar Math Help Forum Discussions

  1. showing something odd for an operator
    Posted in the Differential Geometry Forum
    Replies: 4
    Last Post: Jan 16th 2012, 12:11 PM
  2. Showing existence of 2 constants
    Posted in the Number Theory Forum
    Replies: 7
    Last Post: Nov 18th 2010, 08:41 AM
  3. Showing the existence of solutions
    Posted in the Calculus Forum
    Replies: 13
    Last Post: Jan 18th 2010, 11:57 AM
  4. Showing the existence of two points
    Posted in the Differential Geometry Forum
    Replies: 2
    Last Post: Dec 27th 2009, 12:08 PM
  5. Proving the existence of a unitary operator...
    Posted in the Advanced Algebra Forum
    Replies: 1
    Last Post: Jun 11th 2009, 05:15 PM

Search Tags

/mathhelpforum @mathhelpforum